Benefit cliffs can make a raise cost a single mother real money

Here is a proposition that sounds like a joke and is not. A woman raising two children on a modest paycheck is offered a raise. She does the arithmetic on her whole household budget, taxes and benefits together, and discovers the raise would leave her with less money than she has now. A rational person turns it down. We have built a system that makes the right choice for her family the wrong choice for her future.

This post is about that system. It is a numbers argument, so I will define the terms, build a clearly labelled hypothetical household, show every step of the arithmetic, weigh the strongest objection, and end with design fixes that keep the safety net and remove the trap.

Terms

Means-tested benefit. A program whose payment depends on income. As income rises, the benefit shrinks or disappears. Examples include the Earned Income Tax Credit, the Supplemental Nutrition Assistance Program (SNAP), Medicaid, and federal housing assistance.

Phase-out. The range of income over which a benefit declines. A benefit that falls by 21 cents for each additional dollar of earnings has a phase-out rate of 21 percent.

Effective marginal tax rate (EMTR). The share of an additional dollar of earnings that is lost to higher taxes and lower benefits combined. If you earn $100 more and keep $40, your EMTR is 60 percent. The word "tax" is doing slightly loose work here, since lost benefits are not taxes in a legal sense. In the household's budget, they behave exactly like taxes.

Cliff. A benefit that does not shrink gradually but ends at a threshold, so that one more dollar of earnings can cost thousands of dollars of benefits. A cliff is an EMTR over 100 percent at one point.

The household, stated honestly

Everything in this section is a hypothetical. The programs are real, but the household and the phase-out rates are simplified and chosen for round arithmetic. Real rules vary by state, by household size, by deductions and by year, and most households receive only some of these programs. The purpose is to show the mechanism, not to forecast any actual family's check.

Meet a made-up single parent with two children. Call her Dana. She earns $30,000 a year and receives, in this example, a refundable earned income credit, SNAP, subsidized housing and a health subsidy. She is offered a raise of $10,000, taking her to $40,000. What happens to the extra $10,000?

Add the losses: 765 plus 1,000 is 1,765. Add 2,100 to get 3,865. Add 2,400 to get 6,265. Add 3,000 to get 9,265. Add 1,500 to get 10,765. Subtract that from the $10,000 raise and the household is $765 worse off. The effective marginal tax rate on the raise is 10,765 divided by 10,000, or 107.65 percent.

Each rule in that stack was written separately, by a different committee, for a decent reason. Nobody sat down and designed a 107 percent rate. It is what happens when programs that each seem sensible are stacked on top of one another without anyone adding them up. The household is the only place where they are added up, and she is the one who must live in the total.

Now suppose a smaller version of the same example, with no cliff and a gentler phase-out. Take away the $1,500 health cliff and the sum is $9,265 of loss on $10,000 of raise, or an EMTR of 92.65 percent. Dana keeps $735 of a $10,000 raise. That is no longer a trap in the strict sense, but a raise that is worth 7 cents on the dollar to the person who earns it is a poor incentive to climb.

What CBO and the research say

You do not have to trust my stylized table. The Congressional Budget Office has published analyses of effective marginal tax rates for low- and moderate-income workers. The consistent finding, across those studies, is that the combination of income and payroll taxes with the phase-out of benefits produces high effective rates for many workers in the lower and middle parts of the income distribution, and that the rates vary sharply with a household's mix of programs and with where it sits in the income range. CBO's 2012 report and its 2013 snapshot found that a share of low- and moderate-income workers, over a fifth of them in the snapshot's 2013 estimate, faced combined marginal rates of 40 percent or more, which is at or above the top statutory income tax rate of that year. Look up the agency's reports on effective marginal tax rates for the specific parameters; they publish the assumptions and the household types, and I would rather you read them than take a number from me.

The finding should not be controversial, and it is not a partisan one. A household in the phase-out zone faces a high implicit tax on earning more, whatever your view of how large the safety net should be.

The steelman: the safety net is not designed to be a ladder

The strongest defense of the status quo is worth stating in full.

A benefit that is targeted to the poor must, by arithmetic, decline as income rises. If you want to help only people below a line, then people on the line will lose help as they cross it. The alternative is to pay benefits to everyone, which costs far more, or to cut the benefits, which hurts families who need them. Targeted programs are the most efficient way to deliver help where it is needed. The cliff and the phase-out are not bugs. They are the price of targeting.

That is a real argument and a fair one. It is what the architects of these programs would say. Let me answer it with four points, as a lawyer would, because a verdict requires more than sympathy.

1. The targeting argument proves that some phase-out is necessary, not that this phase-out is wise. A single program that phases out at 21 percent is defensible. Several programs that each phase out at their own rate, in overlapping ranges, with cliffs at their edges, produce a total that no designer chose. The point of a design review is to look at the total. No agency does, because each program belongs to a different department, a different committee and a different appropriation.

1. A cliff is worse than a slope and avoidable. A benefit that ends abruptly at a threshold can be redesigned to taper. The funding cost of tapering is modest compared with the damage of a cliff, because the cliff pays full benefits to people just under the line and none just over it, which is the least efficient shape for targeting.

1. The cost is not only financial. The safety net exists to move people toward self-support. If a worker who accepts a raise ends up with a smaller total income, the program has defeated its own purpose. The trap also builds a quiet habit: it teaches people to keep their hours low, turn down promotions and avoid marriage, since a second earner's income counts in the household total. Reasonable people can disagree about how large those effects are. It is impossible to argue that they point in the right direction.

1. The moral frame matters. The tradition conservatives defend, rooted in the Judeo-Christian view of work as a good and a calling, says that people are made to contribute and to rise. A support system that charges a premium for rising has quietly reversed that logic. Charity that asks a person to remain poor in order to be helped is not charity. It is a managed dependency, however kindly meant.

Why the rules stack: nobody owns the total

Some of this is simple accident. The EITC lives in the tax code and is run by the Treasury. SNAP is a food program under the Department of Agriculture. Medicaid is a joint federal-state program run by state agencies under federal rules. Housing assistance comes from the Department of Housing and Urban Development and local housing authorities. Each has its own definition of income, its own reporting period and its own appeals process. A family that gets help from all four deals with four bureaucracies and four sets of rules for what counts. The Constitution's design of separate powers and federalism has virtues, but this is the dark side of divided authority: when many hands hold the same family, no hand holds the whole picture.

Madison wrote in Federalist No. 62 that laws should not be "so voluminous that they cannot be read, or so incoherent that they cannot be understood." It is in his discussion of the dangers of unstable and unreadable legislation, and anyone who has tried to read the eligibility rules for four benefit programs at once will see what he meant. A rule that a household cannot understand cannot guide the household's choices, and a rule that a household understands only after it has been caught by it is a hidden tax.

The marginal rate at the top, for comparison

It helps to hold two numbers side by side. The top statutory federal income tax rate applies to the highest slice of the highest earners. It is a number that appears in headlines and campaign speeches. The marginal rate on a worker in the phase-out zone is a number that appears nowhere, because no agency publishes the combined rate, and yet in our hypothetical it is higher. The point is not that the top rate should be raised or lowered. The point is that if we truly believed in fairness, we would be as concerned about the rate on the next dollar earned by a mother at $30,000 as about the rate on the next dollar earned by a hedge fund manager at $3,000,000. For why the next dollar's rate matters to behavior, see the marginal rate is the one that decides whether you work the extra shift. For who bears the income tax in the first place, see Who actually pays the federal income tax, read straight from the IRS tables.

Fixes that keep the net and remove the trap

I do not want to leave you with the diagnosis and no treatment. Here are the designs worth fighting for, each with its price.

• Smooth the phase-outs. Replace cliffs with gradual tapers, so that no single dollar of earnings triggers a loss larger than the dollar. This costs some money, since more people stay on a benefit for a longer income range. It buys a continuous incentive to work more.
• Count the total. Require the Treasury and the Office of Management and Budget to publish a combined effective marginal rate for representative households, the way the government already publishes poverty thresholds. What gets measured gets managed.
• Design benefits around work. The Earned Income Tax Credit, for all its flaws, pays more as a worker earns more at the start of the schedule, which rewards the first steps up. A work-based design aims at the same thing for other programs: help grows when hours grow, at least at the bottom. We examine its history and its trouble with errors in The EITC began as a conservative idea and now pays out billions in error.
• Give states room to experiment. Federalism is not a slogan. The states differ in how they handle housing, health coverage and child care, and a state that smooths its own cliffs can show the rest of the country what works. This is what Justice Brandeis meant by laboratories of democracy, and it is the kind of reform that does not need an act of Congress to start.
• Pair benefits with a plain-language statement. A family should be able to see, before accepting a raise, what happens to every benefit. Today the answer often comes in the mail after the fact.

A more ambitious route is to consolidate. Some conservatives have long proposed folding several programs into a single block with a single phase-out, administered by states under a work requirement. Others prefer to keep programs separate but coordinate them. Reasonable people differ, and the evidence on which design works best is mixed. What both camps share is the premise that the system should reward the climb.

The verdict

State it plainly. A safety net that makes a raise unprofitable is not compassionate. It is arithmetic that nobody did, applied to people who cannot afford to be wrong. The families caught in it are not gaming anything. They are doing what the numbers tell them to do. The failure belongs to the designers, who stacked rules without adding them up, and to a political system that has little reason to fix what no committee owns.

Conservatives have a particular obligation here. We say we believe in work, mobility and the dignity of earning one's way. A system that quietly penalizes all three needs our reform energy, not our indifference and not our defensiveness.

The case in brief

1. Means-tested benefits phase out as income rises, and when several phase out together the effective marginal tax rate on the next dollar can approach or exceed 100 percent, as the labelled Dana example shows (107.65 percent with a cliff, 92.65 percent without).
2. CBO's analyses of effective marginal tax rates for low- and moderate-income workers find that combined taxes and phase-outs impose high implicit rates on many households, higher than the top statutory rate for some.
3. Targeting justifies some phase-out but not stacked programs and cliffs that nobody designed as a whole, and the cost falls on people trying to climb.
4. The fixes are smoother tapers, a published combined rate, work-based design and state experiments, each of which keeps help for those who need it.
5. A safety net that charges a premium for rising has reversed its own purpose, and reform should restore the reward for work.

What you can do: ask your state legislator or county benefits office for the income thresholds at which each program you or a neighbor uses begins to phase out, and ask whether anyone publishes the combined effect.